SPATIAL BAYESIAN ESTIMATION OF HIV INFECTION AMONG FEMALE SEX WORKERS (FSW) IN INDIA

The HIV epidemic in India is highly heterogeneous and varies within regions. HIV sentinel surveillance and behavioural surveillance in India are conducted periodically among high-risk groups (HRGs). The data from the surveillances facilitate the estimation of levels and trends of HIV prevalence at the national and state levels. As we aim to achieve the UNAIDS' 'End of AIDS by 2030' goal, HIV prevalence estimations at district and further micro-levels become critical for effective HIV management. However, HIV prevalence data at micro -levels is not available for all districts in India, as the surveillance is being conducted in a limited number of sites. Hence, we propose a Geographical Information System (GIS) based spatial kriging method combined with Bayesian estimation to predict the HIV prevalence among female sex workers (FSW) in India. We used the HIV prevalence of FSWs obtained from the national Integrated Bio-behavioural Surveillance (IBBS) conducted across India in 2014-15. After ensuring spatial autocorrelation, the model prediction and Bayesian estimation were done. A spatial prediction map with the posterior distribution for the model parameters was generated. HIV prevalence in the posterior distribution ranged from 0.5% to 5%, with a higher frequency density in the range of 1% to 3%. The spatial prediction map can be used to estimate the HIV prevalence at micro -levels. It was observed that regions of higher HIV prevalence among FSW were more confined to certain states such as Maharashtra, Karnataka, Andhra Pradesh, Telangana, Rajasthan, Mizoram and Manipur. These estimates will provide invaluable insights on the heterogeneity of HIV prevalence in each state and, thus, will be instrumental in devising region-specific targeted interventions.